Kernel Calculus and Extension of Contact Transformations to D-Modules
Andrea D’Agnolo Pierre Schapira
1 Introduction
There is an important literature dealing with integral transformations. In our papers [6], [7] we proposed a general framework to the study of such transforms in the language of sheaves and D-modules. In particular, we showed that there are two natural adjunction formulas which split many dif- ficulties into two totally different kind of problems: one of analytical nature, the calculation of the transform of a D-module, the other one topological, the calculation of the transform of a constructible sheaf. Similar adjunction formulas for temperate and formal cohomology are obtained by M. Kashi- wara and P. S. in [15], and allow one to treat C
∞-functions and distributions in this framework.
Here, we shall first recall the four above mentioned adjunction formu- las, and then concentrate our study on the D-module theoretical transform.
Given two complex manifolds X and Y and a D-module kernel which defines a quantized contact transformation on an open subset of ˙ T
∗(X × Y ), our main result (Theorem 3.6 below) gives a geometrical condition to extend it as an isomorphism of locally free D-modules of rank one. This improves our previous result of [7].
These results apply to classical problems of integral geometry, in the line of Leray [16], Martineau [18], Gelfand-Gindikin-Graev [9], Helgason [10] or Penrose [8]. In particular, they allowed us to treat projective duality and the twistor correspondence. (Refer to [6], [7], [5]. See also [17] for other flag correspondences.)
Dedicated to Professor Hikosaburo Komatsu AMS classification: 30E20, 32L25, 58G37
Appeared in: “New trends in microlocal analysis” (Tokyo, 1995) Springer-Verlag,
1997, 179–190.
2 Review on the calculus of kernels
In this section we will develop the formalism of kernels in the framework of sheaves and D-modules. The results below concerning kernels for sheaves and D-modules are well-known from the specialists: let us mention in particular M. Kashiwara and also J.-P. Schneiders, with whom we had many discussions on this subject.
2.1 A review on sheaves, D-modules and temperate cohomology
References are made to [14] for the theory of sheaves, and to [19] and [11] for the theory of D-modules (see [22] for a detailed exposition).
Let X be a real analytic manifold, and denote by a
Xthe map from X to the set consisting of a single element. We denote by D
b(C
X) the derived category of the category of bounded complexes of sheaves of C-vector spaces on a topological space X. If A ⊂ X is a locally closed subset, we denote by C
Athe sheaf on X which is the constant sheaf on A with stalk C, and zero on X \ A. We consider the “six operations” of sheaf theory RHom(·, ·),
· ⊗ ·, Rf
!, Rf
∗, f
−1, f
!, and we denote by × the exterior tensor product.
Recall that RHom(·, ·) = Ra
X ∗RHom(·, ·). For F ∈ D
b(C
X) we set D
0F = RHom(F, C
X), DF = RHom(F, ω
M), where ω
X' or
X[dim
RX ] denotes the dualizing complex, and or
Xthe orientation sheaf.
We denote by SS(F ) the micro-support of F , a closed conic involutive subset of T
∗X. We denote by D
bR−c(C
X) the full triangulated subcategory of D
b(C
X) of objects with R-constructible cohomology. If X is a complex manifold, one defines similarly the category D
bC−c(C
X) of C-constructible objects.
Let X be a complex manifold of dimension d
X. We denote by O
Xthe structural sheaf, by Ω
Xthe sheaf of holomorphic forms of maximal degree, and by D
Xthe sheaf of rings of linear differential operators. We denote by Mod(D
X) the category of left D
X-modules, and by Mod
good(D
X) the full sub- category of Mod(D
X) consisting of good D
X-modules. This is the smallest thick subcategory of Mod(D
X) containing the coherent modules which can be endowed with good filtrations on a neighborhood of any compact subset of X. Note that in the algebraic case, coherent D-modules are good. We denote by D
b(D
X) the derived category of the category of bounded com- plexes of left D
X-modules, and by D
bgood(D
X) its full triangulated subcat- egory whose objects have cohomology groups belonging to Mod
good(D
X).
We consider the operations in the derived category of (left or right) D- modules: RHom
DX(·, ·), · ⊗
OLX
·, f
−1, f
!, f
∗. In particular, if M ∈ D
b(D
X),
N ∈ D
b(D
Y), and f : Y −→ X:
f
−1M = D
Y →X⊗
fL−1DXf
−1M, f
!
N = Rf
!(D
X←Y⊗
DLY
N ),
where D
Y →Xand D
X←Yare the transfer bimodules associated to f . We denote by × the exterior tensor product, and we also use the notation:
D
XM = RHom
DX(M, K
X),
where K
Xdenotes the dualizing complex for left D
X-modules, defined by K
X= D
X⊗
OXΩ
⊗−1X[d
X]. If F is a holomorphic vector bundle on X, we set:
F
∗= Hom
OX(F , O
X), DF = D
X⊗
OXF.
Let us briefly recall some constructions of [12] and [15].
First, assume X is a real analytic manifold. Denote by Db
Xthe sheaf of Schwartz’s distributions on X, and by C
X∞the sheaf of functions of class C
∞. There exist unique contravariant functors, exact for the natural t-structures:
T hom(·, Db
X) : D
bR−c(C
X)
op−→ D
b(D
X),
· ⊗ C
w X∞: D
bR−c(C
X) −→ D
b(D
X), such that if Z is a closed subanalytic subset of X, then
T hom(C
Z, Db
X) = Γ
ZDb
X, C
X\Z⊗ C
w X∞= I
Z,X∞,
where I
Z,X∞denotes the ideal of C
X∞of functions vanishing to infinite order on Z.
Now, assume that X is a complex manifold. Denote by X the associated anti-holomorphic manifold, by X
Rthe underlying real analytic manifold, and identify X
Rto the diagonal of X × X. For F ∈ D
bR−c(C
X) one sets:
T hom(F, O
X) = RHom
DX
(O
X, T hom(F, Db
XR)), F ⊗ O
w X= RHom
DX
(O
X, F ⊗ C
w X∞R).
In other words, one defines T hom(F, O
X) and F ⊗ O
w Xas the Dolbeault complexes with coefficients in T hom(F, Db
XR), and F ⊗ C
w X∞Rrespectively.
If F ∈ D
bC−c(C
X), then T hom(F, O
X) has regular holonomic cohomology groups (this is the way Kashiwara proves the Riemann-Hilbert equivalence of categories). In such a case, one has:
T hom(D
0F, O
X) ' D T hom(F, O
X).
If Z is a closed complex submanifold of codimension d of X, we shall con- sider the holonomic left D
X-module B
Z|X= T hom(C
Z[−d], O
X) of [19]. Re- call that B
Z|X' H
[Z]d(O
X) (algebraic cohomology) is a subsheaf of B
Z|X∞= H
Zd(O
X).
2.2 Kernels for sheaves
Here, all manifolds and morphisms of manifolds will be complex analytic.
Let X and Y be complex manifolds of dimension d
Xand d
Yrespectively.
Denote by r : X × Y −→ Y × X the map r(x, y) = (y, x), and by q
1, q
2the first and second projection from X × Y to the corresponding factor. If Z is another manifold, for i, j = 1, 2, 3 we denote by q
ijthe projections from X × Y × Z to the corresponding factor (e.g., q
23: X × Y × Z −→ Y × Z).
Definition 2.1. For K ∈ D
b(C
X×Y) and L ∈ D
b(C
Y ×Z), we set:
K ◦ L = Rq
13!(q
12−1K ⊗ q
23−1L ),
t
K = r
∗D
0K.
Note that the operation ◦ is associative.
For K ∈ D
b(C
X×Y), L ∈ D
b(C
Y ×Z), consider the hypotheses:
(supp(K) × Z) ∩ (X × supp(L)) is proper over X × Z, (2.1) (SS(K) × T
Z∗Z) ∩ (T
X∗X × SS(L)) ⊂ T
X×Y ×Z∗(X × Y × Z). (2.2) Proposition 2.2. Let K ∈ D
bR−c(C
X×Y), L ∈ D
b(C
Y ×Z), and H ∈ D
b(C
X×Z).
Assume (2.1), (2.2). Then:
RHom(H, K ◦ L) ' RHom((
tK) ◦ H, L)[−2d
X].
Proof. One has the chain of isomorphisms:
RHom(H, K ◦ L) = RHom(H, Rq
13!(q
−112K ⊗ q
23−1L )) ' RHom(H, Rq
13∗(q
12−1K ⊗ q
23−1L )) ' RHom(q
13−1H, q
12−1K ⊗ q
−123L)
' RHom(q
13−1H, RHom(q
−112D
0K, q
23−1L)) ' RHom(q
13−1H ⊗ q
12−1D
0K, q
23−1L)
' RHom(q
13−1H ⊗ q
12−1D
0K, q
23!L)[−2d
X] ' RHom(Rq
23!(q
−113H ⊗ q
12−1D
0K[2d
X]), L)
= RHom((
tK ) ◦ H, L)[−2d
X].
Here, we used hypothesis (2.1) in the first isomorphism, and hypothesis (2.2)
in the third.
Corollary 2.3. Let K ∈ D
bR−c(C
X×Y), and assume supp(K) is proper over X,
SS (K) ∩ (T
X∗X × T
∗Y ) ⊂ T
X×Y∗(X × Y ).
Then there are natural morphisms:
C
∆X
−→ K ◦
tK [2d
X], K ◦
tK [2d
Y] −→ C
∆X.
Proof. Applying Proposition 2.2 for Z = X, H = C
∆X, L =
tK we obtain the first morphism. Choosing instead Z = Y , L = C
∆Y, H = K, we get a morphism
tK ◦ K[2d
X] −→ C
∆Y, from which the second morphism in the statement is easily deduced.
Assuming (2.1), (2.2), and assuming that K or L is R-constructible, one proves similarly that
D
0(K ◦ L) ' D
0K ◦ D
0L [2d
Y].
2.3 Kernels for D-modules
Definition 2.4. For K ∈ D
b(D
X×Y) and L ∈ D
b(D
Y ×Z), we set:
K ◦ L = q
13!(q
12−1K ⊗
OLX ×Y ×Z
q
23−1L) Note that
K ◦ L ' q
13!δ
−1(K × L),
where δ denotes the diagonal embedding X × Y × Z −→ X × Y × Y × Z.
Proposition 2.5.
1Let K ∈ D
b(D
X×Y) and N ∈ D
b(D
Y). Then, there is a natural isomorphism in D
b(D
X):
K ◦ N ' Rq
1!(K
(0,dY)⊗
qL−12 DY
q
−12N ), where K
(0,dY)= K ⊗
q−12 OY
q
−12Ω
Yis endowed with its natural (q
1−1D
X, q
2−1D
Y)- bimodule structure.
1
As pointed out to us by Andrei Baran, Proposition B.5 of [7] holds only in the algebraic
case. In the analytic case, it should be replaced by Proposition 2.5 above.
Proof. By definition,
K ◦ N ' Rq
1!((D
X× Ω
Y) ⊗
DLX ×Y
(K ⊗
OLX ×Y
(O
X× N ))) Moreover, one has the following chain of isomorphisms:
(D
X× Ω
Y) ⊗
DLX ×Y
(K ⊗
OLX ×Y
(O
X× N )) ' q
2−1Ω
Y⊗
qL−12 DY
(K ⊗
qL−12 OY
q
2−1N ) ' (q
2−1Ω
Y⊗
qL−12 OY
K) ⊗
qL−12 DY
q
−12N ' K
(0,dY)⊗
qL−12 DY
q
2−1N .
Proposition 2.6. Let K ∈ D
bC−c(C
X×Y), L ∈ D
bC−c(C
Y ×Z), and assume (2.1). Then there is a natural isomorphism:
T hom(K, O
X×Y) ◦ T hom(L, O
Y ×Z) −→ T hom(K ◦ L, O
∼ X×Z)[−d
Y].
Proof. Consider the chain of isomorphisms:
T hom(K, O
X×Y) ◦ T hom(L, O
Y ×Z)
= q
13!(q
12−1T hom(K, O
X×Y) ⊗
OLX ×Y ×Z
q
23−1T hom(L, O
Y ×Z)) ' q
13!(T hom(q
12−1K, O
X×Y ×Z) ⊗
OLX ×Y ×Z
T hom(q
−123L, O
X×Y ×Z)) ' q
13!T hom(q
12−1K ⊗ q
−123L, O
X×Y ×Z)
' T hom(Rq
13!(q
12−1K ⊗ q
23−1L ), O
X×Z)[−d
Y]
= T hom(K ◦ L, O
X×Z)[−d
Y].
For the proof of the above isomorphisms, see [12], [1], [15], and [3]. Note that we used hypothesis (2.1) in the third isomorphism.
Proposition 2.7. For K ∈ D
bgood(D
X×Y) and L ∈ D
bgood(D
Y ×Z), consider the analogous hypotheses to (2.1) (2.2):
(supp(K) × Z) ∩ (X × supp(L)) is proper over X × Z, (2.3) (char(K) × T
Z∗Z) ∩ (T
X∗X × char(L)) ⊂ T
X×Y ×Z∗(X × Y × Z). (2.4) Then there is a natural isomorphism:
D(K ◦ L) ' DK ◦ DL. (2.5)
Proof. Under the above assumptions, D commutes to the operations appear- ing in the definition of ◦.
Proposition 2.8. Let K be a regular holonomic D
X×Y-module, and let K be the associated perverse sheaf K = RHom
DX ×Y(K, O
X×Y). Set
t
K = r
∗DK ' T hom(
tK, O
Y ×X).
Assume (2.1) and (2.2) (or equivalently (2.3) and (2.4)) with Z = X and L =
tK (or equivalently L =
tK). Then there are natural morphisms:
K ◦
tK[d
Y− d
X] −→ B
∆X|X×X,
B
∆X|X×X−→ K ◦
tK[d
X− d
Y].
Proof. Applying Corollary 2.3, we get the morphism:
T hom(C
∆X[−d
X], O
X×X) ←− T hom(K ◦
tK[d
Y], O
X×X)[d
Y− d
X].
The first morphism follows by applying Proposition 2.6. The second mor- phism is similarly obtained.
Remark 2.9. Consider a correspondence:
X ←− S
f−→ Y,
g(2.6)
and denote by h : S −→ X ×Y the morphism h = (f, g). It is then immediate to check that for F ∈ D
b(C
X), K ∈ D
b(C
S), one has:
F ◦ (Rh
!K ) ' Rg
!(K ⊗ f
−1F ).
Moreover, assuming h is proper it is easy to check that for M ∈ D
b(D
X), K ∈ D
b(D
S):
M ◦ (h
!K) ' g
!
(K ⊗
OLS
f
−1M).
Recall that, in the particular case when h is a closed embedding, one has h
!O
S' B
S|X×Y.
2.4 Adjunction formulas
Formulas (2.10), (2.11) below appeared in a slightly more particular situation in [6]. Formulas (2.12), (2.13) are due to [15].
For K a regular holonomic D
X×Y-module, set K = RHom
DX ×Y(K, O
X×Y).
Consider the hypotheses:
(supp(M) × Y ) ∩ supp(K) is proper over Y, (2.7)
(char(M) × T
Y∗Y ) ∩ char(K) ⊂ T
X×Y∗(X × Y ), (2.8)
(X × supp(G)) ∩ supp(K) is proper over X. (2.9)
Theorem 2.10.
2(i) Assuming hypotheses (2.7) and (2.8) above, we have isomorphisms:
RΓ
c(X; RHom
DX(M, (K ◦ G) ⊗ O
X))[d
X] (2.10) ' RΓ
c(Y ; RHom
DY(M ◦ K, G ⊗ O
Y)),
RΓ(X; RHom
DX(M, RHom(G ◦
tK, O
X)))[d
X] (2.11) ' RΓ(Y ; RHom
DY(M ◦ K, RHom(G, O
Y)))[2d
Y].
If moreover (2.9) is satisfied, the formulas above hold interchanging Γ and Γ
c.
(ii) Assuming hypotheses (2.7) and (2.9) above, we have an isomorphism:
RΓ(X; RHom
DX(M, (K ◦ G) ⊗ O
w X))[d
X] (2.12) ' RΓ(Y ; RHom
DY(M ◦ K, G ⊗ O
w Y)).
If moreover (2.8) holds, then we have an isomorphism:
RΓ
c(X; RHom
DX(M, T hom(G ◦
tK, O
X)))[d
X] (2.13) ' RΓ
c(Y ; RHom
DY(M ◦ K, T hom(G, O
Y)))[2d
Y].
Under the same hypotheses, formulas (2.12) and (2.13) hold interchanging Γ and Γ
c.
3 Generalized QCTs
3.1 Kernels for E-modules
We recall here some definitions from the theory of E-modules. We refer to [19]
and to [20] for an exposition.
We denote by π
X: T
∗X −→ X the cotangent bundle to X, by ˙π
X: T ˙
∗X −→ X the cotangent bundle with the zero-section removed, and by T
M∗X the conormal bundle to a submanifold M of X. For a subset V of T
∗X , we set ˙ V = V ∩ ˙ T
∗X . We denote by p
1and p
2the first and second projection from T
∗(X × Y ) ' T
∗X × T
∗Y to the corresponding factor, and by p
a2the composite of p
2with the antipodal map of T
∗Y .
Let E
Xdenote the sheaf of microdifferential operators of finite order on T
∗X. We denote by D
b(E
X) the derived category of the category of bounded complexes of left E
X-modules.
2
In formulas (C.4) and (C.7) of [7], K ◦ G should be replaced by G ◦
tK as in formulas
(2.11) and (2.13) above.
To f : Y −→ X one associates the natural maps:
T
∗Y ←−
tf0
Y ×
XT
∗X −→
fπ
T
∗X.
We will denote by f
µand f
µ
the inverse and direct images in the sense of E-modules. Hence, for M ∈ D
b(E
X) and N ∈ D
b(E
Y):
f
µM = R
tf
0∗(E
Y →X⊗
fL−1π EX
f
π−1M), f
µ
N = Rf
π∗(E
X←Y⊗
tLf0 −1EYt
f
0−1N ), where E
Y →Xand E
X←Yare the transfer bimodules.
If M ∈ D
b(D
X), we set:
EM = E
X⊗
π−1X DX
π
X−1M,
an object of D
b(E
X), and if F is a holomorphic vector bundle on X, we set:
EF = E(DF).
If Z is a closed complex submanifold of codimension d of X, we shall consider the holonomic left E
X-module C
Z|X= EB
Z|X.
Definition 3.1. Let K ∈ D
b(D
X×Y) and L ∈ D
b(D
Y ×Z). Denoting by δ : X × Y × Z −→ X × Y × Y × Z the diagonal embedding, we set:
EK ◦
µEL = q
13µδ
µ(EK × EL).
Proposition 3.2. (i) Let M ∈ D
bgood(D
X), and assume that f : Y −→ X is non-characteristic for M. Then
Ef
−1M ' f
µEM.
(ii) Let N ∈ D
bgood(D
Y). Assume that f is proper on supp N . Then Ef
∗N ' f
µ
EN . (iii) Let M ∈ D
bgood(D
X), N ∈ D
bgood(D
Y). Then
E(M × N ) ' (EM) × (EN ).
Proof. Assertion (iii) is obvious, and (ii) is proved in [21]. Assertion (i)
follows from the division theorem of [19], using the techniques of that paper.
From Proposition 3.2 it immediately follows:
Corollary 3.3. Assuming (2.3), (2.4), there is a natural isomorphism:
E(K ◦ L) ' EK ◦
µEL.
Theorem 3.4. Let K ∈ D
bgood(D
X×Y), N ∈ D
bgood(D
Y). Assume:
supp(K) is proper over X,
char(K) ∩ (T
X∗X × T
∗Y ) ⊂ T
X×Y∗(X × Y ).
Then there is a natural isomorphism:
EK ◦
µEN −→ Rp
∼ 1∗(EK
(0,dY)⊗
pLa 2−1EY
p
a2−1EN ).
Proof. By Corollary 3.3, the left hand side is isomorphic to E(K ◦ N ). Hence, by Proposition 2.5, it is enough to prove the natural isomorphism:
E
XRq
1!(K ⊗
DLY
N ) −→ Rp
∼ 1!(E
X×YK ⊗
DLY
N ).
Let E
X×Y /Ydenote the subsheaf of E
X×Yof sections which commute with O
Y. The second hypothesis, and the division theorem of [19] gives:
E
X×Y /YK ' E
X×YK
Let K
0be a coherent O
X×Y-module which generates K, and N
0be a coherent O
Y-module which generates N . Let L
0= K
0⊗
OYN
0. It is enough to check:
E
X(0) ⊗
OXRp
1!L
0−→ Rp
∼ 1!(E
X×Y /Y(0) ⊗
OX ×YL
0). (3.1) E
X(0) is an O
X-module of type DFN and E
X×Y /Y(0) ' E
X(0) b ⊗O
Y(see [21, I §7]). Then the proof goes as that of Theorem 7.3 of [15], using Proposition 3.13 of [21] (or Theorem 8.1 of [15]).
3.2 Extending QCTs to D-modules
In this section, we assume X and Y have the same dimension n. Let L be a regular holonomic D
X×Y-module, and set:
L = RHom
DX ×Y(L, O
X×Y), Λ = char(L) ⊂ T
∗(X × Y ).
Let F and G be holomorphic line bundles on X and Y respectively, and set:
L
(n,0)(F , G) = q
1−1(F ⊗
OXΩ
X) ⊗
q−11 OX
L ⊗
q−12 OY
q
2−1G.
Lemma 3.5. Assuming q
2is proper on supp L, there is a natural isomor- phism:
α : Γ(X × Y ; L
(n,0)(F , G
∗)) ' Hom
Db(DY)(DG, DF ◦ L). (3.2) Proof. It is enough to apply H
0(·) to the following chain of isomorphisms:
Ra
X×Y ∗L
(n,0)(F , G
∗) ' ' Ra
Y ∗Rq
2∗RHom
q−12 DY
(q
−12DG, L
(n,0)⊗
q−11 OX
q
1−1F) ' Ra
Y ∗RHom
DY(DG, Rq
2∗(L
(n,0)⊗
q−11 OX
q
1−1F)) ' Ra
Y ∗RHom
DY(DG, DF ◦ L).
Here, in the first isomorphism we used the fact that DG is D
Y-coherent, and in the last one we used the fact that q
2is proper on supp L.
Assuming q
2is proper on supp L, by Lemma 3.5 we associate to s ∈ Γ(X × Y ; L
(n,0)(F , G
∗)) the D
Y-linear morphism:
α (s) : DG −→ DF ◦ L.
Let U
Xand U
Ybe two open conic subsets of ˙ T
∗X and ˙ T
∗Y respectively.
Set
Λ
0= Λ ∩ (U
X× U
Ya), Σ = Λ \ Λ
0, W = T
∗Y \ U
Y. We assume:
(a) ˙ W is a C-analytic closed conic subset of ˙ T
∗Y of codimension ≥ 2, (b) Λ
0is a smooth Lagrangian manifold, and p
1: Λ
0−→ U
X, p
a2: Λ
0−→
U
Yare isomorphisms (in other words, Λ
0defines a contact transforma- tion),
(c) p
a−12W ˙ ∩ ˙Λ = ˙Σ, and this analytic set has dimension < n, (d) L has no submodules isomorphic to O
X×Y,
(e) there exists s ∈ Γ(X × Y ; L
(n,0)(F , G
∗)), which is non-degenerate on Λ
0.
We refer to [19] for the notion of non-degenerate section of a holonomic module.
Theorem 3.6. Assume hypotheses (a)–(e) above. Then:
H
0(α(s)) : DG −→ H
0(DF ◦ L)
is an isomorphism.
Proof. Applying Corollary 3.3, it is enough to prove the isomorphism EG −→
∼H
0(EF ◦
µEL) all over T
∗Y . Consider the morphism of distinguished trian- gles:
RΓ
W(EG) −→ Rπ
Y ∗(EG) −→ RΓ
UY(EG) −→
+1↓
αW(s)↓
α(s)↓
αUY(s)RΓ
W(EF ◦
µEL) −→ Rπ
Y ∗(EF ◦
µEL) −→ RΓ
UY(EF ◦
µEL) −→
+1(3.3) By (b) and (c), RΓ
UY(EF ◦
µEL) ' RΓ
UY(EF |
UX◦
µEL|
Λ0). Since s is non-degenerate on Λ
0, α
UY(s) is an isomorphism by [19]. Applying H
0(·), we get the commutative diagram in which the horizontal lines are exact:
H
W0EG → DG → H
U0YEG −→ H
W1EG
↓ ↓
H0(α(s))↓
o↓
H
W0(EF ◦
µEL) → H
0(DF ◦ L) → H
U0Y(EF ◦
µEL) −→ H
β W1(EF ◦
µEL).
We shall prove (i) H
W0EG = 0,
(ii) H
W0(EF ◦
µEL) = 0, (iii) H
W1EG = 0.
This will imply the result, for then β will be the zero morphism.
Since the problem is local on T
∗Y , in (i) and (iii) we may assume G = O
Y. Then (i) and (iii) follow from [13, Theorem 1.2.2]. To prove (ii), consider the isomorphisms:
RΓ
W(EF ◦
µEL) ' RΓ
WRp
a2 ∗(EL
(n,0)⊗
p−11 OX
p
−11F) ' Rp
a2∗RΓ
Σ(EL
(n,0)⊗
p−11 OX